Consent banners vs reality: how to test what your website does before a regulator does
A banner is a promise made in words. The test is whether the network traffic keeps it. Anyone with a browser can run the test, and so can a regulator or a plaintiff.
Key points
- A consent banner test compares what the banner and privacy notice say with what the site actually loads and sends before any choice, after a rejection, after acceptance and when an opt-out signal is present.
- The most common failure is a rejected choice that is recorded while the trackers still fire, usually because a tag was added directly to the page and bypasses the consent tool.
- Repeat the test after every release that touches tags, the consent tool or the tag manager, because a new marketing pixel is added far more often than the banner is changed.
The test needs no special tools: a clean browser profile, the developer tools’ network and storage panels, and a written record of what you saw. Findings are facts about the site. What the law requires of those facts is for your counsel.
Five test runs, in order
Each run uses a fresh browser profile with no history, so earlier visits cannot hide a problem. The order matters because the first run sets the baseline the rest are compared with.
Run one: before any choice
Open a clean profile, open the network panel, and load the home page without touching the banner. Save the list of requests and the cookies and local storage entries. Group them by the party they go to: your own domain, your hosting, and everyone else. Third parties that appear before a choice need an explanation. Some are needed to deliver the page or keep it secure; others, such as advertising and analytics tags, usually depend on the visitor’s choice under the rules that apply to you.
Check also what the page does at load that is not a cookie: a pixel request that carries a page address, a font or script fetched from a third party, and an embedded video or map.
Run two: reject, then reload
The most common failure is a rejected choice that is recorded while the trackers still fire, usually because a tag was added directly to the page and bypasses the consent tool.
Choose the reject or decline option, reload the page, and navigate to two or three others. Compare the requests with run one. Anything optional that still appears is a finding. Check also that the banner treats reject as easily reachable as accept: if one is on the first layer and the other needs three clicks, record it as a design finding for counsel, who knows how the rule reads in your market.
Then test persistence: does the rejection survive a reload and a return visit, or does the banner reappear every time?
Run three: accept, then compare
Accept all, reload, and list what is added. This is the list the banner’s own description should match. Look for a vendor that fires after acceptance but is not named anywhere, and for categories described in the notice that you cannot find in the traffic. Both directions are findings: one is undisclosed processing, the other is a notice that no longer describes the site.
If the site passes visitor choices to Google tags through consent mode, check that all four consent types, ad_storage, analytics_storage, ad_user_data and ad_personalization, change when the choice changes. The types are defined in Google’s Tag Manager help.
Run four: the opt-out signal
Global Privacy Control is a browser or extension setting that signals a visitor’s privacy choice to websites, and California’s regulations require businesses to treat a qualifying opt-out preference signal as a valid opt-out request. The signal is described at globalprivacycontrol.org, and the California regulations are on the California Privacy Protection Agency’s site.
To test it, enable the setting in a browser that supports it, load the site, and look at the traffic as in run two. The optional advertising and sharing tags should stay quiet. How the signal applies to a signed-in account, and whether other laws require the same, is a question for counsel.
Run five: everywhere else
Repeat the main steps on a landing page built by marketing, a checkout or sign-up page, a logged-in page, a microsite on another domain, and the mobile web version. These are where tags are added by someone who was not told about the banner. Mobile apps are a separate test, because they use software development kits instead of cookies.
Repeat the test after every release that touches tags, the consent tool or the tag manager, because a new marketing pixel is added far more often than the banner is changed.
Reading the results
| What you see | What it usually means |
|---|---|
| Optional tags fire before any choice | Tags hard-coded in the page or tag manager, outside the consent tool |
| Tags still fire after rejection | Tag rules not tied to the consent categories, or a second tag manager |
| A vendor fires but is not in the notice | A tag added without review |
| Signal ignored | No code reads the signal, or it is read only in one region |
The ICO finalised its guidance on storage and access technologies on 29 April 2026, after consulting on a draft that reflected changes to the UK’s e-privacy rules made by the Data (Use and Access) Act, so earlier summaries of the UK position may be out of date. The ICO’s guidance is on its storage and access technologies page. Consent and the rules around cookies differ by country, and for the GDPR the lawful basis for each purpose is a decision for your counsel. For the Canadian standard of consent, see meaningful consent under PIPEDA and BC PIPA.
What to do with the findings
Rank each finding by what leaves the site, to whom, and about whom. Fix the technical causes first, because they are cheap. Update the notice to match, then repeat the test and keep the before and after captures as evidence. The service that does this, including the review of the tag manager and the consent tool, is website consent and cookie review.
How the work is bounded
The scope is agreed in writing before work starts, and the engagement is quoted in writing with it.
SecHB does not issue certifications, attestations or audit opinions: those come from accredited certification bodies, CPA firms and QSAs. The work here is what an organization does to be ready for them.
Nothing here is legal advice. Where a question turns on the law, the work is done alongside the client’s counsel, not instead of them.
Questions we are asked
What is a consent banner test?
A consent banner test compares what the banner and privacy notice say with what the site actually loads and sends before any choice, after a rejection, after acceptance and when an opt-out signal is present.
What is the most common failure?
The most common failure is a rejected choice that is recorded while the trackers still fire, usually because a tag was added directly to the page and bypasses the consent tool.
How should opt-out signals be tested?
Global Privacy Control is a browser or extension setting that signals a visitor’s privacy choice to websites, and California’s regulations require businesses to treat a qualifying opt-out preference signal as a valid opt-out request.
Has UK guidance changed recently?
The ICO finalised its guidance on storage and access technologies on 29 April 2026, after consulting on a draft that reflected changes to the UK’s e-privacy rules made by the Data (Use and Access) Act, so earlier summaries of the UK position may be out of date.
How often should we repeat the test?
Repeat the test after every release that touches tags, the consent tool or the tag manager, because a new marketing pixel is added far more often than the banner is changed.
See what the site does
Send the address of the site and the regions your visitors come from. The reply says what a review would cover. The service is website consent and cookie review.